Repository navigation
Conversation
Both polarisation columns of the cumulative phase come from identical stored terms (now asserted exactly), but delta_phi contracts them in one np.tensordot, and the BLAS kernel may round each output column differently: on an AVX-512 build with OpenBLAS 7 and 19 of 1536 elements differed in the last bit. Compare the columns within 4 gamma_n(u) of sum |w| |t| per element (testing.md section 3, class R). The bound was spot-checked: a 1e-13 relative perturbation of one column fails it by 12 to 17 times. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XGENPwwkUUG9KSAwneHjgH
The device filter piped grep -vE into grep -qE. On a long change list with the native match first, grep -q exits at that match, the upstream grep dies of SIGPIPE, and under pipefail the negated pipeline sets device=false despite a native change. The second grep now reads the whole pipe and discards its output. The current step runs as bash -e, so this hardens against a shell-policy change rather than fixing a skipped lane. tests/test_ci_changes.py runs the workflow's own step under bash -e and bash -e -o pipefail: a 100 000-line list with the native match first (fails under pipefail before this change), and the existing positive and negative classifications, including an unknown base. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XGENPwwkUUG9KSAwneHjgH
jtravs
marked this pull request as ready for review
November 13, 2026 19:29
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Documentation build overview
21 files changed ·
|
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
gamma(2n) * sum|w||t|(about 4 gamma_n(u)). Their stored terms are identical, and that is now asserted exactly, but the increment is one BLAStensordotwhose kernel may round output columns differently (reproduced on an AVX-512 build with OpenBLAS: 19 of 1536 elements differ in the last bit). A 1e-13 relative perturbation fails the bound by 12–17×.changesjob's device filter reads the whole pipe (grep -E … > /dev/nullinstead ofgrep -qE), sopipefailcannot turn a native change intodevice=falsethrough SIGPIPE. A regression test (tests/test_ci_changes.py) extracts the workflow's own step and runs it underbash -eandbash -e -o pipefail, with a 100,000-line list and the existing classifications.Not included: #777. The object-file pass it proposes finds nothing under the default IPO/LTO Release build; that needs a build-policy decision.
Closes #494
Closes #612
🤖 Generated with Claude Code
https://claude.ai/code/session_01XGENPwwkUUG9KSAwneHjgH